Mapping Quantitative Trait Loci for Yield Potential Traits in Wheat Recombinant Inbred Lines

: Selecting wheat with favorable spike characteristics has been a popular breeding strategy to improve sink capacity and yield potential. In the present study, quantitative trait loci (QTLs) for yield potential traits were identiﬁed using a recombinant inbred line (RIL) population derived from the cross between Taejoong and Keumkang, two Korean wheat cultivars. A linkage map encompassing a total genetic length of 6544.8 cM was constructed using 838 single nucleotide polymorphisms from the 35K Axiom Wheat Breeder’s Array. We detected eight QTLs for four yield potential traits that are consistently identiﬁed in at least two of the three environments, that is, one for days to heading date ( QDHD-1 on chromosome 7B), three for spike length ( QSL-1 , QSL-2 , and QSL-3 on chromosomes 1D, 5A, and 6A, respectively), one for tiller number ( QTN-1 on chromosome 5B), and three for length of center rachis ( QLCR-1 , QLCR-2 , and QLCR-3 on chromosomes 1B, 5B, and 6A, respectively). Notably, Taejoong contributed the alleles for long spike at all three spike length QTLs with the additive effects of 0.6 cm, 0.6 cm, and 0.9 cm at QSL-1 , QSL-2 , and QSL-3 , respectively. No signiﬁcant two-way or three-way interaction was observed among QSL-1 , QSL-2 , and QSL-3 , indicating that pyramiding the Taejoong alleles at the three QTLs can increase spike length additively. While the Taejoong alleles at QSL-1 , QSL-2 , and QSL-3 were associated with increased days to heading date, more kernels per spike, and reduced tiller number per plant, the extent of the pleiotropic effects were different among the three QTLs. Due to the limited number of molecular markers and mapping resolution, further work is required to narrow down the identiﬁed QTLs and characterize their effects more precisely. Our results would provide useful information for modulating spike characteristics and improving yield potential in wheat breeding programs. Data curation, C.-S.K. and Y.-J.M.; Formal analysis, K.-M.K., K.-H.K., and J.-B.C.; Validation, C.-S.P.; Writing—original draft, C.-S.K. and Y.-J.M.; Writing—review & editing, C.-S.P.


Introduction
Improving grain yield is an important goal in wheat breeding to secure stable food production under changing climate and increasing world population [1,2]. Wheat grain yield is a complex trait involving a number of genes and their interactions which are largely influenced by various environmental factors including temperature, photoperiod, rainfall, and fertilization [3,4]. Therefore, studies on wheat yield generally focus on specific yield component traits such as spikes (or tillers) per unit area, kernels per spike, and kernel weight that are more stably inherited than the final grain yield [5,6]. While studies genetically dissecting a specific yield component trait can provide useful molecular tools for breeding programs, optimizing the balance between different yield components is also Agronomy 2021, 11,22 2 of 14 critical as the yield component traits are interrelated and an increase in one component is frequently associated with a decrease in another, i.e., the yield improvement effect of an increase in kernels per spike can be canceled out by a decrease in kernel weight [7][8][9][10][11].
In many wheat breeding programs, kernels per spike has been a major target for improving yield as it often shows stronger positive correlation with the final grain yield compared with other yield components [12]. For example, the yield increasing trend of the major Argentinian wheat cultivars released between 1920 and 1990 was associated better with the increases in kernels per spike compared with those in spikes per unit area or grain weight [13]. Similarly, over 400 European winter wheat cultivars released since the 1950s showed continuous increases in kernels per spike over time [14]. At the International Maize and Wheat Improvement Center (CIMMYT), large spike wheat cultivars were developed by wide crossing between common wheat and accessions of Agropyron elongatum or Triticum polonicum to improve yield potential through increasing kernels per spike [15]. While these lines showed significant increases in both kernels per spike and kernel weight, the final grain yield fell short of expectations mainly due to the decrease in spikes per unit area [16,17].
In South Korea, a large spike wheat cultivar Taejoong was developed from the cross between the Chinese wheat Xian83 and the Korean cultivar Keumkang at the National Institute of Crop Science (NICS) [18,19]. In the regional yield trials conducted for three years (2014)(2015)(2016) at the NICS, spike length and kernels per spike of Taejoong were 13.4 cm and 48, respectively, which were 76% longer and 66% more than those of Keumkang, respectively [19]. Taejoong also exhibited a ten-thousand kernel weight of 48.3 g, which was 5% heavier than Keumkang. Although Taejoong exhibited limited tillering capacity, i.e., 417 spikes per square-meter, which was 42% less than Keumkang, its final grain yield was moderately higher (13-21% higher depending on the cropping patterns) than Keumkang, mainly due to the superior spike characteristics in terms of kernel number and size [19]. As the genetic factors underlying the large spike characteristics of Taejoong remain unclear, we aimed to identify quantitative trait loci (QTLs) for yield potential characteristics of Taejoong using a recombinant inbred line (RIL) population derived from the cross between Taejoong and Keumkang and generate useful information for improving yield potential in wheat breeding programs.

Plant Materials
A total of 94 F 10 RILs were produced from the cross between two Korean wheat cultivars, Taejoong and Keumkang. The main phenotypic feature of Taejoong is a large spike ( Figure 1) [19,20]. The RIL population was cultivated in upland conditions at Jeonbuk National University (35 •

Measurement of Yield Potential Traits
Major agronomic traits affecting yield potential were recorded as follows: days to heading date (DHD) as the number of days from sowing to spike exsertion from the stem in 50% of a plot, culm length (CL) as the length from ground level to the base of the spike, spike length (SL) as the length of the spike excluding awn, number of tillers (TN) counted as the number of stems per plant, kernels per spike (KS) counted as the number of grains per spike, length of center rachis (LCR) as measured excluding flowerets, and the value of compactness (COM) as calculated according to the number of kernels per spike divided by spike length. CL, SL, and TN were measured from 20 plants randomly selected in each plot at maturity. KS, LCR and COM were determined from the main spikes of the same 20 plants.

Genotypic Analysis
Genomic DNA (gDNA) from the young leaves of the F10 RIL population was extracted using a DNA extraction kit (Solgent, Korea) according to the manufacturer's instructions. Concentration of each gDNA was adjusted to 20 ng/μL with distilled water. The final concentration of gDNA was measured by Biodrop (Biodrop Ltd., Cambridge, UK). The 94 RILs were genotyped by DNA Link, Inc. (Korea, http://www.dnalink.com) with Axiom TM 35K Wheat Breeder's Genotyping Array (ThermoFisherScientific, Applied Biosystems TM , Waltham, US) in a 384-array plate format. Genotype calling was conducted by Affymetrix Power Tools (APT, release 2.11.1).

QTL Mapping and Statistical Analysis
A linkage map was constructed using the IciMapping program version 4.2 [21]. To select polymorphic single nucleotide polymorphisms (SNPs), the genotyping result of the 35,042 SNPs was converted by the "SNP" function of the program. Out of 35,042 SNPs, 2061 were polymorphic between Taejoong and Keumkang, the parental lines of the RIL population. Using the "BIN" function in IciMapping, SNPs with more than 10% missing data and/or significant segregation distortion (Chi squared test P-value < 0.001) were deleted. Finally, 838 SNPs were used to construct a linkage map by the "MAP" function. QTL mapping for yield potential traits (DHD, CL, SL, COM, KS, TN, and LCR) was performed by composite interval mapping. Significant QTLs were initially identified based on the threshold logarithm of odds (LOD) score of 3.0 in each year, and those consistently identified in at least two of the three years were declared as stable QTLs. Statistical analyses including ANOVA and correlation analysis were conducted using R (The R Project for Statistical Computing version 3.4.4, https://www.r-project.org).

Measurement of Yield Potential Traits
Major agronomic traits affecting yield potential were recorded as follows: days to heading date (DHD) as the number of days from sowing to spike exsertion from the stem in 50% of a plot, culm length (CL) as the length from ground level to the base of the spike, spike length (SL) as the length of the spike excluding awn, number of tillers (TN) counted as the number of stems per plant, kernels per spike (KS) counted as the number of grains per spike, length of center rachis (LCR) as measured excluding flowerets, and the value of compactness (COM) as calculated according to the number of kernels per spike divided by spike length. CL, SL, and TN were measured from 20 plants randomly selected in each plot at maturity. KS, LCR and COM were determined from the main spikes of the same 20 plants.

Genotypic Analysis
Genomic DNA (gDNA) from the young leaves of the F 10 RIL population was extracted using a DNA extraction kit (Solgent, Daejeon, Korea) according to the manufacturer's instructions. Concentration of each gDNA was adjusted to 20 ng/µL with distilled water. The final concentration of gDNA was measured by Biodrop (Biodrop Ltd., Cambridge, UK). The 94 RILs were genotyped by DNA Link, Inc. (Seoul, Korea, http://www.dnalink.com) with Axiom TM 35K Wheat Breeder's Genotyping Array (ThermoFisherScientific, Applied Biosystems TM , Waltham, MA, USA) in a 384-array plate format. Genotype calling was conducted by Affymetrix Power Tools (APT, release 2.11.1).

QTL Mapping and Statistical Analysis
A linkage map was constructed using the IciMapping program version 4.2 [21]. To select polymorphic single nucleotide polymorphisms (SNPs), the genotyping result of the 35,042 SNPs was converted by the "SNP" function of the program. Out of 35,042 SNPs, 2061 were polymorphic between Taejoong and Keumkang, the parental lines of the RIL population. Using the "BIN" function in IciMapping, SNPs with more than 10% missing data and/or significant segregation distortion (Chi squared test P-value < 0.001) were deleted. Finally, 838 SNPs were used to construct a linkage map by the "MAP" function. QTL mapping for yield potential traits (DHD, CL, SL, COM, KS, TN, and LCR) was performed by composite interval mapping. Significant QTLs were initially identified based on the threshold logarithm of odds (LOD) score of 3.0 in each year, and those consistently identified in at least two of the three years were declared as stable QTLs. Statistical analyses including ANOVA and correlation analysis were conducted using R (The R Project for Statistical Computing version 3.4.4, https://www.r-project.org).

Phenotype Evaluation of the RIL Population
Descriptive statistics of the yield potential traits of the Taejoong × Keumkang RILs are summarized in Figure 2 and Table S1. The average DHD (184.9 days) of the RILs was shorter than Taejoong (192.7 days) and longer than Keumkang (180.3 days), and its distribution was skewed toward late DHD. The distribution of CL was normal and the average (85.3 cm) was taller than Keumkang (66.1 cm) and similar to Taejoong (84.6 cm). The average SL (11.5 cm) was shorter than Taejoong (14.5 cm) and longer than Keumkang (8.4 cm), and its distribution was skewed toward short SL. The distribution of COM was normal and the average (4.5) was higher than both Keumkang (4.4) and Taejoong (4.2). The distribution of KS was normal and the average (50.1) was between Keumkang (36.7) and Taejoong (61.6). The average TN (6.1) was greater than Taejoong (5.1) and fewer than Keumkang (8.1), and its distribution was skewed toward a low value. The average LCR (0.68 cm) was shorter than Taejoong (0.75 cm) and similar to Keumkang (0.69 cm), and its distribution was skewed toward a low value.
Of 21 correlation pairs among the seven yield potential traits, nine were significant in all three years, 2017, 2018, and 2019 ( Figure 3 and Figures S1-S3). Strong positive correlations were observed between SL and LCR (r = 0.69 in all three years). KS showed moderate to low levels of positive correlations with COM (r = 0.45-0.47), SL (0.34-0.40), and DHD (r = 0.26-0.30). Relatively strong negative correlations were observed between COM and SL (r = −0.62 to −0.57), and COM and LCR (r = −0.54 to −0.50). TN was negatively, but weakly correlated with SL (r = −0.37 to −0.33), KS (r = −0.33 to −0.29), and LCR (r = −0.28 to −0.25). CL was not correlated with any of the six other yield potential traits, except for a weak positive correlation with SL (r = 0.21) significant in only one of the three years ( Figure S3).

QTLs for Yield Potential Traits
A linkage map spanning a total genetic length of 6544.8 cM was constructed using 838 SNPs selected from the 2061 polymorphic SNPs after excluding those with >10% missing data and/or significant segregation distortion (Chi squared test p-value < 0.001, Table S2). The number of SNPs polymorphic between Taejoong and Keumkang was relatively small (5.9% of the 35K SNPs) mainly due to the fact that Taejoong had been derived from the cross between Xian83 and Keumkang and shared similar genetic background with Keumkang.
Of the 838 SNPs used for mapping, 233 (27.8%) were distributed on the A genome, 344 (41.1%) were on the B genome, and 261 (31.3%) were on the D genome. Among the 838 SNP markers, chromosome 1B had the greatest number of markers (86), whereas chromosome 6A had the smallest number of markers (9). Chromosome 5B showed the longest length (578.4 cM), whereas chromosome 4D showed the shortest length (113.1 cM). The average length between two adjacent markers were the longest on chromosome 6A (19.5 cM) and the shortest on chromosome 2B (4.1 cM).   We detected eight QTLs for four yield potential traits that are consistently identified in at least two of the three years (Table 1 and Figure 4). One QTL for DHD, QDHD-1 on  We detected eight QTLs for four yield potential traits that are consistently identified in at least two of the three years (Table 1 and

Effects of QSL-1, QSL-2, and QSL-3 on Spike Length and Other Yield Potential Traits
To evaluate the main effects of the three QTLs for SL and their interactions on yield potential traits, three-way factorial ANOVAs of QSL-1, QSL-2, and QSL-3 were carried out for each trait with the year (2017, 2018, and 2019) as a random factor ( Table 2). Spike length (SL): The main effects of QSL-1, QSL-2, and QSL-3 were highly significant for SL. Compared to the RILs carrying the Keumkang allele at each QTL, those carrying the Taejoong allele exhibited significantly longer spike, i.e., 1.2, 1.2, and 1.8 cm longer for QSL-1, QSL-2, and QSL-3, respectively. Notably, there was no significant two-way or three-way interaction among the three QTLs, indicating that pyramiding of the Taejoong alleles at QSL-1, QSL-2, and QSL-3 would increase SL additively. While the average SL of the RILs carrying the Keumkang alleles at all three QTLs was 9.7 cm, those carrying the Taejoong allele at one of the three QTLs exhibited the average SL of 10.6-11.1 cm ( Figure 5). The RILs carrying the Taejoong alleles at two of the three QTLs exhibited the average SL of 11.8-12.5 cm. The RILs carrying the Taejoong alleles at all three QTLs exhibited the longest average SL of 13.8 cm. Days to heading date (DHD): QSL-1 and QSL-2 had significant main effects on DHD, with the Taejoong alleles for long spike being associated with late heading. The RILs carrying the Taejoong allele at QSL-1 headed 1.0 day later than those carrying the Keumkang allele. Similarly, the RILs carrying the Taejoong allele at QSL-2 headed 0.8 days later than those carrying the Keumkang allele. The main effect of QSL-3 was not significant for DHD. While there was no significant two-way interaction, the three-way interaction was significant and explained 4.7% of the DHD variance. Culm length (CL): While the main effects of QSL-1 and QSL-2 were not significant for CL, that of QSL-3 was highly significant. The RILs carrying the Taejoong allele at QSL-3 were 3.9 cm taller than those carrying the Keumkang allele. The QSL-2 × QSL-3 interaction and the three-way interaction were significant for CL, but explained limited levels (3.0% and 2.3%, respectively) of the CL variance. Compactness (COM): The main effects of QSL-1, QSL-2, and QSL-3 were highly significant for COM, with the RILs carrying the Keumkang allele exhibiting more compact spike than those carrying the Taejoong allele. No significant two-way or three-way interaction was observed. Kernels per spike (KS): QSL-1 and QSL-3 had significant main effects on KS. Compared to the RILs carrying the Keumkang allele at QSL-1, those carrying the Taejoong allele had 2.8 more kernels per spike. Similarly, those carrying the Taejoong allele at QSL-3 had 2.8 more kernels per spike than those carrying the Keumkang allele. Two-way or three-way interactions were not significant, indicating that the Taejoong alleles at QSL-1 and QSL-3 would increase KS additively. Tiller number (TN): The main effects of QSL-1, QSL-2, and QSL-3 were significant for TN, with the RILs carrying the Keumkang allele exhibiting more tillers than those carrying the Taejoong allele. The QSL-1 × QSL-2 interaction and the QSL-2 × QSL-3 interaction were significant for TN, explaining the limited level (3.6% and 1.6%, respectively) of the TN variance. Length of center rachis (LCR): The main effects of QSL-1, QSL-2, and QSL-3 were highly significant for LCR, with the RILs carrying the Taejoong allele exhibiting longer LCR than those carrying the Keumkang allele. The QSL-1 × QSL-2 interaction and the QSL-1 × QSL-3 interaction were significant, explaining the limited level (2.0% and 1.4%, respectively) of the LCR variance.

Potential Use of the Spike Length QTLs in Breeding
Improving yield potential has been one of the most important goals in wheat breeding [22][23][24][25]. As kernels per spike is strongly correlated with the final grain yield, selecting breeding lines with large spike has been a popular strategy for increasing sink capacity and improving yield potential [12,26]. To genetically dissect the large spike characteristics of the Korean wheat cultivar Taejoong, we identified three major QTLs for spike length, QSL-1, QSL-2, and QSL-3, using F10 RILs which were derived from the Taejoong × Keumkang cross and genotyped with high density SNPs. At all three QTLs, the Taejoong alleles for long spike were associated with greater number of kernels per spike. Therefore, the SNPs linked to QSL-1, QSL-2, and QSL-3 would provide useful molecular tools to develop breeding lines with increased sink capacity through marker assisted selections. QSL-1, QSL-2, and QSL-3 are especially promising as they exhibited no significant twoway or three-way interactions on spike length and kernels per spike, indicating that pyramiding the Taejoong alleles at these QTLs would additively increase sink capacity (Table 2, Figure 5).
Nevertheless, caution is required as the effect of increased kernels per spike is frequently compensated for by decreased tiller number per unit area [27][28][29][30]. Our results showed that the Taejoong alleles for long spike at QSL-1, QSL-2, and QSL-3 are also associated with reduced tiller number, and that the three QTLs exhibit different levels of tradeoffs between kernels per spike and tiller number. For example, while the effect of the Taejoong allele increasing kernels per spike was similar between QSL-1 and QSL-3, the effect of the Taejoong allele decreasing tiller number was nearly seven times greater at

Potential Use of the Spike Length QTLs in Breeding
Improving yield potential has been one of the most important goals in wheat breeding [22][23][24][25]. As kernels per spike is strongly correlated with the final grain yield, selecting breeding lines with large spike has been a popular strategy for increasing sink capacity and improving yield potential [12,26]. To genetically dissect the large spike characteristics of the Korean wheat cultivar Taejoong, we identified three major QTLs for spike length, QSL-1, QSL-2, and QSL-3, using F 10 RILs which were derived from the Taejoong × Keumkang cross and genotyped with high density SNPs. At all three QTLs, the Taejoong alleles for long spike were associated with greater number of kernels per spike. Therefore, the SNPs linked to QSL-1, QSL-2, and QSL-3 would provide useful molecular tools to develop breeding lines with increased sink capacity through marker assisted selections. QSL-1, QSL-2, and QSL-3 are especially promising as they exhibited no significant two-way or three-way interactions on spike length and kernels per spike, indicating that pyramiding the Taejoong alleles at these QTLs would additively increase sink capacity (Table 2, Figure 5).
Nevertheless, caution is required as the effect of increased kernels per spike is frequently compensated for by decreased tiller number per unit area [27][28][29][30]. Our results showed that the Taejoong alleles for long spike at QSL-1, QSL-2, and QSL-3 are also associated with reduced tiller number, and that the three QTLs exhibit different levels of tradeoffs between kernels per spike and tiller number. For example, while the effect of the Taejoong allele increasing kernels per spike was similar between QSL-1 and QSL-3, the effect of the Taejoong allele decreasing tiller number was nearly seven times greater at QSL-1 compared with that at QSL-3 (Table 2). While further investigation is required to characterize the effects of the spike length QTLs on other yield potential traits under different environments and genetic backgrounds, the current results suggest that QSL-3 is likely more beneficial than QSL-1 in terms of optimizing the balance between kernels per spike and tiller number. Our results also illustrate the importance of dissecting individual genetic factors underlying yield potential traits and characterizing their effects on other yield components in order to provide breeding programs with useful molecular tools for optimizing the balance among different yield components.
Supplementary Materials: The following are available online at https://www.mdpi.com/2073-439 5/11/1/22/s1, Figure S1: Distribution and correlation of yield potential traits in 94 F10 recombinant inbred lines produced from the cross between Taejoong and Keumkang in 2017. T and K indicate Taejoong and Keumkang, respectively. M (black dia-mond) indicates the average value of each trait. DHD, days to heading date; CL, culm length; SL, spike length; COM, compactness; KS, kernels per spike; TN, number of tillers; LCR, length of center rachis. An asterisk indicates significance (, *, and *** at P ≤ 0.05, P ≤ 0.01, and P ≤ 0.001 levels, respectively). Figure S2: Distribution and correlation of yield potential traits in 94 F10 recombinant inbred lines produced from the cross between Taejoong and Keumkang in 2018. T and K indicate Taejoong and Keumkang, respectively. M (black dia-mond) indicates the average value of each trait. DHD, days to heading date; CL, culm length; SL, spike length; COM, compactness; KS, kernels per spike; TN, number of tillers; LCR, length of center rachis. An asterisk indicates significance (, *, and *** at P ≤ 0.05, P ≤ 0.01, and P ≤ 0.001 levels, respectively). Figure S3: Distribution and correlation of yield potential traits in 94 F10 recombinant inbred lines produced from the cross between Taejoong and Keumkang in 2019. T and K indicate Taejoong and Keumkang, respectively. M (black dia-mond) indicates the average value of each trait. DHD, days to heading date; CL, culm length; SL, spike length; COM, compactness; KS, kernels per spike; TN, number of tillers; LCR, length of center rachis. An asterisk indicates significance (, *, and *** at P ≤ 0.05, P ≤ 0.01, and P ≤ 0.001 levels, respectively). Table S1: Descriptive statistics of the yield potential traits in the Taejoong × Keumkang recombinant inbred linesa. Table S2: SNP-based genetic map for QTLs related to yield potential traits in 94 F10 recombinant inbred lines produced from the cross between Taejoong and Keumkang.